Hi Dave, You've got two different things happening here.
> -----Original Message----- > From: Dave Kliczbor [mailto:[EMAIL PROTECTED] > Sent: Wednesday, September 03, 2003 12:52 PM > To: [EMAIL PROTECTED] > Subject: [SAtalk] Bayes filter and autolearning > > > | X-Spam-Status: No, hits=4.1 required=5.0 > | tests=BAYES_90,NO_REAL_NAME > | autolearn=ham version=2.55 > > auto_learn_threshold_nonspam is set to 1.9. > Now I am wondering why on earth SpamAssassin learns that message as > _ham_? As far as I can see, this should not happen. On this one, it's because autolearn learns the message pre-Bayes test. Without your Bayes_90, that message scored under 1.9 points, and so it was autolearned. The other reason that applies here is explained below. > On the other hand, I have more than one mail where > SpamAssassin behaves > like that: > > | X-Spam-Status: Yes, hits=11.8 required=5.0 > | tests=BAYES_70,FORGED_MUA_OUTLOOK,MSG_ID_ADDED_BY_MTA_3, > | NIGERIAN_BODY,RATWARE_OE_MALFORMED,RISK_FREE > | version=2.55 > > auto_learn_threshold_spam is set to 5.0. > Why doesn't SpamAssassin learns this message as spam? > There's a "safety zone" around your spam hits in 2.55. It's not there in 2.60. If you run spamassassin -D --lint you'll see a line like: debug: auto-learn? safety=4, ham=1.9, spam=5, body-hits=-0.4, head-hits=-1.3 This means that the only spams you'll auto-learn as spam are ones that are NOT within 4 points of your "spam level" setting. So in your case, you simply cannot learn spams <9 and you simply cannot learn hams >1. No matter what you set your auto_learn_thresholds to. -tom ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk